bumie-e/whisper-small-lora-afrispeech

Whisper Small Yoruba - Bunmi Akinremi

This model is a fine-tuned version of openai/whisper-small on the Afrispeech dataset. It achieves the following results on the evaluation set:

• Loss: 0.9628
• Wer: 31.5147

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9701	1.1136	1000	1.0837	34.1506
0.9156	2.2272	2000	1.0030	32.5710
0.8785	3.3408	3000	0.9774	31.9314
0.8716	4.4543	4000	0.9660	31.6407
0.9619	5.5679	5000	0.9628	31.5147

Framework versions

• PEFT 0.18.2.dev0
• Transformers 5.3.0
• Pytorch 2.10.0+cu128
• Datasets 3.6.0
• Tokenizers 0.22.2